K.O. In The Final Round: Stability Stops After 750 MHz FSB
While our benchmarks should be familiar to you, the results require a little explanation. For starters, we weren't able to reach our goal of stable 800 MHz FSB. The system stopped being more or less stable after 750 MHz. With patience and some crashes in between, we also achieved the performance level at 800 MHz FSB.
This time, though, we didn't swap the processor, preferring to stick with the same model and its set multiplier (2.26 GHz with a x17 multiplier). That means that each uptick in FSB clock pumped up the processor clock and the RAM clock. The only clock speed we left was for PCI and AGP - at 33 and 66 MHz, respectively; the BIOS on the Asus P4G8X permits these settings. Since all clock speeds increased proportionately, we obtained wonderfully scalable results.
The following table explains the results in the benchmark diagrams:
| CPU | FSB Clock | Eff. CPU Clock | Memory Clock | AGP Clock |
|---|---|---|---|---|
| P4 2.26 | 533 MHz | 2.26 GHz | 266 MHz | 66 MHz |
| P4 2.26 | 600 MHz | 2.55 GHz | 300 MHz | 66 MHz |
| P4 2.26 | 666 MHz | 2.83 GHz | 333 MHz | 66 MHz |
| P4 2.26 | 700 MHZ | 2.98 GHz | 350 MHz | 66 MHz |
| P4 2.26 | 750 MHz | 3.20 GHz | 376 MHz | 66 MHz |
| P4 2.26 | 800 MHz | 3.40 GHz | 400 MHz | 66 MHz |
| P4 3.06 | 533 MHz | 3.06 GHz | 266 MHz | 66 MHz |
| P4 3.06 | 592 MHz | 3.40 GHz | 298 MHz | 66 MHz |
We included a Pentium 4 3.06 as a control - to ensure comparability, we took one without HyperThreading. We also used this CPU to reach the 3.4 GHz touted by Intel the old-fashioned way: by overclocking the FSB from 133 to 148 MHz.
